Séminaires

Test

The Centre for Systems Engineering and Applied Mechanics (CESAME) organizes a series of seminars. The seminars are held in the CESAME lecture room, Building EULER, 4-6, av. Georges Lemaître, Louvain-la-Neuve (Parking 13).

For further information, feel free to contact the secretary: 010/47.80.36.

If you wish to receive the seminar announcements by email, please send an email to Lydia De Boeck.

Look also at the seminars of the departement of mecanics and at the seminars of the research group on Large Graphs and Networks.

Seminars to come

23/03/2010 (14.00) [Lieu : Bât. Euler]
Olivier BONAVENTURE (UCL-INGI)
"One path is not enough anymore"
The Internet was designed with the assumption that there was one path between a pair of communicating hosts. The TCP congestion control scheme was added in the late 1980s to avoid congestion collapse on bottleneck links. The growth of multihoming, both at the enterprise/campus network level and at the host level with multiple wireless interfaces, completely changes the situation since there are now several useable paths between communicating hosts. In this talk, I'll survey the broad implications of this change on the network, transport and application layers of the Future Internet.


6/04/2010 (14.00) [Lieu : Bât. Euler]
Samih ZEIN
"Polynomial Chaos Expansion in Reliability Analysis and Robust Optimisation"
The Polynomial Chaos Expansion (PCE) is an efficient numerical method for relating the output of a nonlinear system with the uncertainty in its input parameters. The output can be approximated with a multidimensional polynomial function using the so called PCE. Numerically, such approximation can be obtained using a linear regression with an adequate design of experiments. In this presentation, a review is given on the PCE principle and its application in reliability analysis and robust optimisation. Then, an optimization algorithm is proposed that seeks the best design of experiments in the D-optimal sens for the PCE. This algorithm is based on the heuristic optimisation and the Fedorov exchange algorithm and is adapted to the specific case of the polynomial chaos expansion. We will conclude the presentation with a brief overview of the integration of a PCE method into an industrial optimization platform and task manager."


13/04/2010 (14.00) [Lieu : Bât. Euler]
Emilie MARCHANDISE (CESAME)
"Quality meshing of medical geometries with harmonic map"
This work describes an automatic approach to recover a high quality surface mesh from low-quality or oversampled inputs (STL-files) obtained from medical imaging through classical segmentation techniques. The approach combines a robust method of parametrization based on harmonic maps with a recursive call to a multi-level edge partitioning software. By doing so, we are able to get rid of the topological and geometrical limitations of harmonic maps. The overall remeshing procedure is implemented, together with the ïnfinite element discretization procedure required for computing harmonic maps, in the open-source mesh generator Gmsh. We show that the proposed method produces high quality meshes and we highlight the benefits of using those high quality meshes for biomedical simulations.


20/04/2010 (14.00) [Lieu : Bât. Euler]
Guillaume LECLERCQ (CESAME)
"The brain accounts for 3D eye and head kinematics in the visuomotor transformation of velocity signals in manual tracking"
Visually guided arm movements are common actions carried out each day in our everyday life. When performing visually guided arm movements, the brain needs to transform the visual information (the tennis ball, an obstacle) into an appropriate motor plan for the arm (i.e. a command to give to the arm’s muscles). Here, we investigated if the brain accounts for the 3D eye position, 3D head position and 3D eye velocity in the visuomotor transformation of velocity signals for manual tracking movements. A visuomotor transformation model taking into account the complete 3D eye-head-shoulder kinematics was built using dual quaternions. To test model predictions, human subjects performed manual tracking movements in different eye-head-shoulder configurations, either static or dynamic. We compared the arm initial open-loop direction with the predictions provided by the model. Results show that the brain uses internal 3D eye and head position as well as eye velocity signals to account for the eye-head-shoulder kinematics in the visuomotor transformation of velocity signals for manual tracking movements.


27/04/2010 (14.00) [Lieu : Bât. Euler]
Didier HENRION (LAAS-CNRS, University of Toulouse )
"Occupation measures and LMI formulation of piecewise affine optimal control design problems"
After recalling the mathematical framework of occupation measures for nonlinear dynamical systems control, we show how linear matrix inequalities (LMI) can be used to solve computationally the corresponding generalized problem of moments. We illustrate these ideas first on a simple linear quadratic regulator design problem, and then on a less trivial piecewise affine optimal control problem for which we know analytically the solution of the Hamilton-Jacobi-Bellman equation. This is joint work with Luis Rodrigues, Concordia University, Montreal, Canada.







Previous seminars

02/03/2010 (14.00) [Lieu : Bât. Euler]
Giovanni SAMAEY (CESAME)
Numerical properties of equation-free methods for multiscale problems
In quite a number of problems, there exists a wide gap between the (microscopic) level at which a system is modeled and the (macroscopic) space and time scales at which one would like to observe the system. Equation-free methods try to bridge this gap by constructing a coarse time-stepper for the macroscopic observables as a three-step procedure: (1) lift (from macro to micro); (2) evolve (using the microscopic model); and (3) restrict (from micro to macro). We will discuss both how to construct lifting operators to obtain a well-defined coarse time-stepper, as well as what macroscopic tasks can be performed once a coarse time-stepper is available. Particular emphasis will be put on the numerical properties of the resulting methods (accuracy and efficiency) and on the use of approximate macroscopic models to accelerate the computations. Collaborators that have contributed to this work will be acknowledged throughout the talk.


23/02/2010 (14.00) [Lieu : Bât. Euler]
Manuel MARTINEZ (University of Seville )
"Stability of Asynchronous Feedback-Interconnected Dissipative Systems"
This work presents a method to analyze the stability of the feedback interconnection of a class of dissipative linear systems when the sampling associated to the feedback interconnection is asynchronous. We consider systems that are either Input/Output Strictly passive (IOSP), or systems which have bounded L2-gains less than one. The analysis is performed by using the concept of MAximum Sampling time preserving Dissipation (MASD), for each interconnected system. We investigate the impact of using the scattering transformation in the computation of the MASD, and we provide a numerical algorithm (based on a set of LMI's) that allows to choose the most suitable configuration for the interconnection.


18/02/2010 (16.30) [Lieu : Auditoire Ch.-J. de La Vallée Poussin (cycl 01) ]
Jean-Charles DELVENNE (FUNDP Namur)
"Colloquium CESAME-MAPA "Random walks on graphs: what for?"
How does Google work? What is so magical about expander graph? How do we detect communities in large social graphs? The three answers involve one same tool: random walks on graphs. Given a graph, one may start from a vertex and, at each time step, jump randomly to one of its neighbours. This random walk transforms the graph into a Markov chain. The study of this Markov chain provides a lot of insight into the combinatorial properties of the graph. We will cover a sample of applications of this technique.


11/02/2010 (16.30) [Lieu : Auditoire Euler]
Pierre BIELIAVSKY (UCL)
"Colloquium CESAME-MAPA "Géométrie non commutative et théorie de Lie." (slides)
Mon but est d'exposer de manière élémentaire les concepts à la base de la géométrie non commutative d'Alain Connes et, si le temps le permet, de montrer certaines implications en géométrie différentielle et en théorie de Lie. L'ensemble C(X) des fonctions continues à valeurs complexes sur un espace topologique X a naturellement une structure d'algèbre associative commutative. En effet, étant données deux fonctions f:X-->C et g:X-->C, leur produit est la fonction f.g définie par (f.g)(x)=f(x).g(x). Dans de nombreux cas, cette structure d'algèbre encode complètement la structure topologique de l'espace X. Ce résultat classique, dû à I. Guelfan'd, fournit un outil algébrique pour l'étude des espaces topologiques. Dans les années 80, Alain Connes introduit la géométrie différentielle non commutative basée sur l'idée que des situations géométriques apparemment disparates, incluant la géométrie riemannienne et la théorie des feuilletages, peuvent elles aussi être encodées par une notion algébrique commune: celle de triplet spectral. Dans mon exposé, j'introduirai les paradigmes de bases de cette théorie en me basant sur un concept émergeant de la théorie quantique: la notion de déformation.


09/02/2010 (14.00) [Lieu : Bât. Euler]
Mariya ISHTEVA (CESAME)
"Numerical methods for the best low multilinear rank approximation of higher-order tensors". (slides)
Higher-order tensors are generalizations of vectors and matrices to third- or even higher-order arrays of numbers. The multilinear rank of a tensor is a direct generalization of column and row rank of a matrix. In this talk we discuss the best low multilinear rank approximation. Given a higher-order tensor, we are looking for another tensor, as close as possible to the original one and with multilinear rank bounded by prespecified numbers. This approximation is used for dimensionality reduction and signal subspace estimation in higher-order statistics, biomedical signal processing, telecommunications and many other fields. We consider two particular applications, present several algorithms for the computation of the approximation and discuss the issue of local minima.


16/12/2009 (16.15) [Lieu : Auditoire Ch.-J. de La Vallée Poussin (cycl 01) ]
Marc BURGER (ULg)
Colloquium CESAME-MAPA "Surfaces, espaces de Teichmüller et groupes de Lie"
Il est bien connu depuis Nielsen que les surfaces (compactes connexes sans bord) sont classifiées, à homéomorphisme près, par leur groupe fondamental. Cette situation change radicalement lorsqu'on adopte le point de vue de la géométrie différentielle: une surface S donnée porte en effet de nombreuses structures Riemanniennes différentes. En fait, à toute métrique hyperbolique sur S on associe une représentation linéaire de son groupe fondamental $\pi_1(S)$ à valeurs dans SL(2,R). On obtient ainsi un ensemble de représentations, appelé "espace de Teichmüller", qui porte une structure très riche. L'objet de cet exposé est de décrire ce qui se produit lorsqu'on remplace SL(2,R) par un groupe de Lie de dimension supérieure.


15/12/2009 (14.00) [Lieu : Bât. Euler]
Quentin RENTMEESTERS (CESAME)
"Filtering on manifolds" (slides)
In many applications related to signal processing and image processing, a filtering technique is required to reduce the influence of perturbations on the measurements. In this talk, we will see different approaches to implement such a filtering technique when the measurements belong to a nonlinear space. We will illustrate these techniques on the sphere and show some concrete applications related to the Grassmann manifold.


08/12/2009 (14.00) [Lieu : Bât. Euler]
Charlotte BEAUTHIER (FUNDP)
"Modelling Competitive Coexistence in a Chemostat by LQ Type State Feedback"
The chemostat model describes the interaction of microbial species which are competing for a single nutrient. It has been used for different systems such as lakes, waste-water treatment processes and biological reactors producing ge- netically altered organisms. The competitive exclusion principle is a theoretical result which states that the competition process yields at best a single winning species in the long run. However, in nature, many species seem to coexist. This contradiction between the theory and the real world is our main motivation for modifying the model in order to try to bring them in better accordance. In this talk, the problem is stated in terms of an LQ-optimal control problem for a linearized system in order to garantee the coexistence of species for the (nonlinear) chemostat model. Conditions are reported such that the computed linear state feedback can be applied to the nonlinear system and garantee the well-posedness of the model and in particular the coexistence of the species. The results are illustrated by some numerical experiments. This contribution is a joint work in progress with Joseph Winkin (FUNDP) and Denis Dochain (UCL).


01/12/2009 (14.00) [Lieu : Bât. Euler]
Damien ERNST (ULg)
" Learning about near-optimal policies from a sample of trajectories" (slides)
In many areas of the medical, social and engineering sciences the following planning problem arises. A training set of trajectories representing the system state, the actions and the instantaneous rewards at successive discrete instants is known and a decision policy leading to high cumulative rewards has to be found. In the first part of this talk, I will present some batch mode reinforcement learning algorithms for solving these types of problems. Afterwards, I will report on some recent work to build lower bounds on the performance of a policy from trajectories. Finally, I will show that this work has opened a frontier in batch mode reinforcement learning whose final implications we cannot yet guess.


24/11/2009 (14.00) [Lieu : Bât. Euler]
Luc VANDENDORPE (ELEC/TELE, UCL)
"Rate-Optimized Power Allocation for Relayed OFDM transmission in different scenarii" (slides)
We consider an OFDM (orthogonal frequency division multiplexing) point-to-point transmission scheme which is enhanced by means of a relay. Symbols sent by the source during a first time slot may be (but are not necessarily) retransmitted by the relay during a second time slot. The relay is assumed to be of the DF (decode-and-forward) type. For each relayed carrier, the destination implements maximum ratio combining. Two protocols are considered. Assuming perfect CSI (channel state information), the paper investigates the power allocation problem so as to maximize the rate offered by the scheme for two types of power constraints. Both cases of sum power constraint and individual power constraints at the source and at the relay are addressed. The theoretical analysis is illustrated through numerical results for the two protocols and both types of constraints. Then the case where multiple relays are active will be considered. Finally a multiple access scheme with 2 users will be discussed.


17/11/2009 (14.00) [Lieu : Bât. Euler]
Eric BULLINGER (ULg)
"A Systems Biology Approach to Apoptosis Signalling"
Apoptosis is an important physiological process crucially involved in the development and homeostasis of multi-cellular organisms. Although the major signalling pathways leading from the extrinsic induction to the execution of apoptosis have been unravelled, a detailed mechanistic understanding of the complex underlying network and the signal crosstalk remains elusive. A systems biology approach allows to combine diverse data into mathematical models to perform predictive simulations and testing of quantitative and dynamical hypotheses. The modelling process furthermore reveals theoretical and computational challenges.


10/11/2009 (14.00) [Lieu : Bât. Euler]
Michel GEVERS (CESAME)
"Identifiability, informativity, information matrix and the Prediction Error criterion: a new look at the connections " (slides)
This seminar will present the results of 3 years of work with Alex Bazanella (Universidade Federal do Rio Grande do Sul), Xavier Bombois (Delft University) and Ljubisa Miskovic (EPFL, Lausanne) on the connections between four different ingredients in Prediction Error Identification: the information content of the data, the identifiability of the model structure, the information matrix, and the minimum (or minima) of the Prediction Error Criterion. We show that by introducing a new concept of local informativity of the data, one can establish a number of nice equivalences between these four ingredients.


03/11/2009 (14.00) [Lieu : Bât. Euler]
Christophe DE VLEESCHOUWER (TELE, UCL)
"Multi-sensored vision for autonomous production of personalized video summaries"
Today's media consumption evolves towards increased user-centric adaptation of contents, to meet the requirements of users having different expectations in terms of story-telling, and heterogeneous constraints in terms of access devices. Individuals and organizations want to access dedicated contents through a personalized service that is able to provide what they are interested in, at the time when they want it, and through the distribution channel of their choice. In this talk, we explain how it is possible to address this challenge by merging computer vision tools with optimization and resource allocation mechanisms, to automate the collection and distribution of audiovisual contents. In a typical application scenario, the sensor network for media acquisition is composed of (microphones and) cameras, which, for example, cover a basket-ball field. Distributed analysis and interpretation of the scene are exploited to decide what to show or not to show about the event, so as to produce a video composed of a valuable subset from the streams provided by each camera. The process involves numerous integrated technologies and methodologies, including but not limited to automatic scene analysis, camera viewpoint selection and control, and generation of summaries through automatic organization of stories. Considering the problem in a multi-camera environment not only mitigates the difficulty of scene understanding caused by reflection, occlusion and shadow in the single view case, but also offers higher flexibility in producing visually pleasant video reports. In final, multi-camera autonomous production/summarization can provide practical solutions to a wide range of applications, such as personalized access to local sport events through a web portal or a mobile hand-set, cost-effective and fully automated production of content dedicated to small-audience, e.g. souvenirs DVD, university lectures, conference, etc, and interactive browsing and automated summarization for video surveillance.


27/10/2009 (14.00) [Lieu : Bât. Euler]
John LEE (Unité d'imagerie moléculaire et radiothérapie expérimentale, UCL)
"Dimensionality reduction: from PCA to recent nonlinear techniques" (slides)
Dimensionality reduction is an old yet unsolved problem, with many applications in data visualization, knowledge discovery, and machine learning in general. Our aim in this talk will be to review several developments in the field of dimensionality reduction, with a particular focus on nonlinear methods. As an introduction, we will point out some weird properties of high dimensional spaces, which will motivate the use of dimensionality reduction. Next, we will go back in time and start our review with a short reminder about well known techniques such as principal component analysis and multidimensional scaling. Our travel into time will also bring us to visit Sammon mapping and other methods based on distance preservation. Next, we will come across self-organizing maps and auto-encoders with bottleneck neural networks. Some spectral methods such as Isomap and locally linear embedding will be reviewed as well. A glance at recent methods based on similarity preservation such as stochastic neighbor embedding will close the survey. Finally, we will try to identify the relationships between the different approaches, and say a few words about quality criteria for dimensionality reduction techniques.


20/10/2009 (14.00) [Lieu : Bât. Euler]
Jan MACIEJOWSKI (University of Cambridge)
"Fault-tolerant control - is it possible?" (slides)
A real demand now exists for fault-tolerant control systems. This talk gives some examples in which this arises, drawn from process control, automotive and aerospace applications. But fault-tolerance is a very ambitious objective; the paper examines critically what has been achieved so far, and what it might be possible to achieve. It also criticises current academic work on fault-tolerant control, and speculates a little on how progress might be achieved. The paper is illustrated by examples from real applications, particularly from the aerospace sector.


13/10/2009 (14.00) [Lieu : Bât. Euler]
Richard BRAATZ (University of Illinois at Urbana-Champaign)
"Robust Optimal Control of Finite-time Distributed Parameter Systems"
Most products of high value such as in the pharmaceuticals, microelectronic, and nanotechnology industries are manufactured in a series of processing steps that operate over finite time. These processes are usually distributed parameter systems in which tight control is required. Computationally efficient methods are proposed for the robust optimal control of finite-time distributed parameter systems (DPS), in which robustness is ensured for either deterministic or stochastic parametric uncertainties. In the deterministic case, the effects of uncertainties on the states and product quality are quantified by power series expansions combined with linear matrix inequality or structured singular value analysis. In the stochastic case, the effects of uncertainties are quantified by power series or polynomial chaos expansions followed by Monte Carlo simulation. The robust performance analysis have been incorporated into fixed or model predictive control algorithms. The approaches are illustrated for several applications problems.


08/10/2009 (16.30) [Lieu : Auditoire Ch.-J. de La Vallée Poussin (cycl 01) ]
Michel WEISS (Aberdeen)
Colloquium CESAME-MAPA "Espaces de surfaces et catégories de surfaces (conjecture de Mumford)"
Soit M l'espace des surfaces connexes, orientées, de genre g, plongées dans l'espace euclidien de dimension n. On suppose que g et n sont très grands. L'espace M est très loin d'être simplement connexe. Toutefois la conjecture de Mumford, dans une formulation contemporaine, est une description de la meilleure approximation de M par un espace simplement connexe. Selon la conjecture celle-ci s'obtient à partir de M par un procédé de "scanning". L'étude de certaines catégories de surfaces a beaucoup contribué à la preuve de la conjecture.


06/10/2009 (14.00) [Lieu : Bât. Euler]
CHANG Chia-Tche , GORISSEN Bastien (CESAME)
"Fast computation on SO(3) of optimal oriented bounding box" (slides)
Photorealistic realtime rendering and physically correct simulations of 3D scenes are at the bleeding edge of what can be done with today's computers. Such applications involve collision detection and visibility tests using methods based on oriented bounding boxes (OBB) enclosing 3D objects. Joseph O'Rourke published in 1985 an algorithm that computes the exact solution in cubic time, but is extremely hard to implement. In practice, faster PCA-based and brute force heuristics are used. The computation of the minimal-volume OBB is formulated as an unconstrained optimization problem on the rotation group SO(3). It is solved using a hybrid method combining the genetic and Nelder-Mead algorithms. This method is analyzed then compared to the current state-of-the-art, and it turns out to be either faster or more accurate.


29/09/2009 (14.00) [Lieu : Bât. Euler]
Pierre DAYE ((CESAME))
"New model of gaze tracking in 2D: Novel architecture with independent gaze and head controllers" (slides)
Reorienting our gaze recruits both eye and head movements to achieve a common goal. During a gaze movement, the head may also be voluntarily moved for some other purpose, e.g., to align the mouth with food. Currently, two mechanisms are proposed to explain the coordination of eye and head movements. In the first, a central gaze controller pre-computes a desired eye and head displacement to execute a gaze movement. This model can generate independent movements for the head and gaze but, because of its structure, it can not compensate for perturbations of the head during the gaze change. In the second, a gaze controller sends the same command to the eye and the head, and a gaze feedback loop compensates for any errors or perturbations during the gaze change. However, this model can not generate independent movements of the head. We propose a novel approach for coordinated eye and head control in which there are separate controllers, for gaze and head, rather than for gaze alone or for eye and head. In this model, the gaze controller drives both the eye and the head. Gaze feedback through the cerebellum compensates for perturbations by acting only on the eye. Thus, in this model the eye only serves gaze, and can not be directed to an arbitrary goal.


22/09/2009 (14.00) [Lieu : Bât. Euler]
Pierre-Antoine , Absil (CESAME)
"Stiefel Manifolds and their Applications" (slides)
Stiefel manifolds, or sets of orthonormal p-frames in R^n, are named after Eduard Stiefel (1909-1978), who analyzed their topology in the 1930s. This presentation will be a reiteration of a talk that was part of the celebration of Stiefel's 100th birthday at ETH Zurich. We will count dimensions, describe tangent spaces, visualize embedded and quotient geodesics, and touch upon applications such as principal component analysis, Lyapunov exponent computation, and blind source separation.


15/09/2009 (14.00) [Lieu : Bât. Euler]
Brian ANDERSON (Australian National University)
"Multivariable zero-free Transfer Functions and Spectra, and their application in Economic Modelling.". (slides)
Central banks and funds investment managers work with mathematical models.  In recent years, a new class of model has come into prominence—generalized dynamic factor models. These are characterized by having a modest number of inputs, corresponding to key economic variables and industry-sector-wide variables for central banks and funds managers respectively, and a large number of outputs, economic time series data or individual stock price movements for example. It is common to postulate that the input variables are linked to the output variables by a finite-dimensional linear time-invariant discrete-time dynamic model, the outputs of which are corrupted by noise to yield the measured data.  The key problems faced by central banks or funds managers are model fitting given the output data (but not the input data), and using the model for prediction purposes. These are essentially tasks usually considered by those practicing identification and time series modelling. Nevertheless there is considerable underlying linear system theory. This flows from the fact that the underlying transfer function matrix is tall.

This presentation will describe a number of consequences of this seemingly trivial fact. For example, a tall transfer function of known McMillan degree but otherwise generic has no zeros, finite or infinite. A finite sequence of output data in the discrete time case allows recovery of a finite sequence of input data, without knowledge of the initial state. Canonical state-variable forms take on a special structure, with the number of real parameters growing linearly with the number of outputs, rather than, as usual, quadratically.


26/08/2009 (16.30) [Lieu : Aud. Ch.-J. de la Vallée Poussin (Cycl01)]
Svetlana BUTLER (University of Illinois at Urbana-Champaign )
Colloquium CESAME-MAPA "Quasi-linear functionals and topological measures"
This talk is an introduction to the theory of quasi-linear functionals and topological measures. Quasi-linear functionals are functionals on C(X) which, rather than being linear on all of C(X), are only required to be linear on singly generated subalgebras of C(X). The corresponding set-functions on X are topological measures. Quasi-linear functionals and topological measures generalize linear functionals and Borel measures and have many fascinating properties. We will give basic definitions and examples, discuss some major results in the field so far, and outline venues for further research in this new area of mathematics.


30/06/2009 (14.00) [Lieu : Bât. Euler]
Raphaël JUNGERS (CESAME )
"The joint spectral radius: theory and applications."
I will survey recent results on the joint spectral radius, and more generally on asymptotic properties of products of matrices. I will put emphasis on nice applications and on open problems, like the finiteness property.


04/06/2009 (16.30) [Lieu : Bât. Euler]
Pierre ROUCHON (Ecole des Mines, Paris)
Colloquium CESAME-MAPA "Fonctions Gevrey et contrôle frontière de certaines EDP"
Après des rappels historiques et définitions sur l'ordre Gevrey des fonctions indéfiniment dérivables, nous montrerons leur intérêt pour la planification de trajectoires de certains systèmes décrits par des équations aux dérivées partielles avec contrôle au bord. Les deux exemples traités seront les suivants: équation de la chaleur avec une frontière libre (problème de Stefan); une particule quantique dans une boite. Nous montrerons d'une part, en quoi ces deux exemples sont représentatifs de la platitude différentielle pour les systèmes de dimension infinie, et d'autre part l'intérêt, encore mal compris, des techniques de re-sommation sur les développements en séries des trajectoires.


26/05/2009 (14.00) [Lieu : Bât. Euler]
Damien FRANCOIS (CESAME)
"Subset selection for prediction analysis of NIR Spectra"
Spectrometric data involve very high-dimensional observations issued from sampled spectra. The correlation of the resulting spectral variables and the large number of them imply several difficulties in modelling. This talk will propose a method for clustering the variables in order to perform subset selection, in order to reduce dimensionality, colliearity, and therefore complexity. Experiments on real world data show that the reduction in redundancy and in number of features leads to better performances obtained using a very low number of spectral ranges.


19/05/2009 (14.00) [Lieu : Bât. Euler]
Ilse IPSEN (North Carolina State University )
"Subset Selection"
Subset selection methods try identify those columns of a matrix that are "most" linearly independent. Many subset selection methods are based on a QR decomposition with column pivoting. We discuss deterministic and probabilistic methods for subset selection, as well as the application of subset selection to nonlinear least squares problems in parameter estimation.


12/05/2009 (14.00) [Lieu : Bât. Euler]
Semiha TURKAY (Institute of Automatic Control, Technical University of Munich )
"Influence of tire damping on the ride performance potential of quarter-car active suspensions"
In this study, constraints on the transfer functions from the road disturbance to the vertical acceleration, the suspension travel, and the tire deflection are derived for a quarter-car active suspension systems using the vertical acceleration and/or the suspension travel measurements for feedback. The influence of tire damping on the design of an active suspension system is illustrated by a mixture of the LQG methodology and the interpolation approach. Besides, H2 and mixed H2/H? syntheses of the quarter-car model excited by random road disturbances are studied. It is demonstrated that the influence of tire damping on the closed-loop performance of the active suspension system can be significant. Finally, the result is extended to suspension models with polytopic tire damping uncertainties


05/05/2009 (14.00) [Lieu : Bât. Euler]
Cristina STOICA (SUPELEC-FR.)
"Robustification hors ligne de lois de commande prédictives multivariables"
Cette présentation propose une méthodologie hors ligne pour la robustification de lois de commande prédictives multivariables, se basant sur une problématique d’optimisation convexe d’un paramètre de Youla. Le point de départ de la démarche consiste à synthétiser une loi de commande initiale prédictive multivariable sous forme d’état qui stabilise le système. Le but est de garantir la robustesse en stabilité face à des incertitudes non structurées et d’assurer des performances nominales pour le rejet de perturbations, imposées sous la forme des gabarits temporels sur les sorties. Ce problème d’optimisation est résolu par un formalisme LMI. Le paramètre de Youla obtenu permet de gérer d’une part le compromis entre la robustesse en stabilité et les performances nominales et d’une autre part permet de réduire l’influence du couplage multivariable sur le rejet des perturbations. Le cas de systèmes incertains appartenant à un ensemble donné d’incertitudes polytopiques est également traité. Dans ce cas, une condition supplémentaire BMI est ajoutée pour chaque sommet du polytope considéré, afin de garantir la stabilité sur tout le polytope. Il s’agit d’un problème d’optimisation non-convexe pour lequel une solution de complexité raisonnable sous une forme LMI sous-optimale est proposée. L’originalité consiste à pouvoir garantir la stabilité avec le correcteur robustifié pour le domaine incertain considéré, même dans le cas où le correcteur MPC initial est instable sur une partie du polytope.


28/04/2009 (14.00) [Lieu : Bât. Euler]
Iven MAREELS (Dept. of Electrical and Electrotechnic Eng., University of Melbourne)
"Water Information Networks: Successes and Challenges" (slides)
It is estimated that we harvest and utilize about 65% of the readily available fresh water resources of the world. In general, perhaps because water is perceived as an abundantly available resource, we use water rather poorly. Typically less than half the water taken from the environment serves the objective for which it was intended. The UNESCO World Water reports 2003 and 2005 identify in no uncertain terms a water crisis.

In this lecture we provide an overview of a 10 year collaborative research and development effort, between the University of Melbourne and a local company Rubicon Systems Australia, and more recently with National ICT Australia. The programme called Water Information Networks (WIN) is a systems engineering approach to water management in irrigation systems. Because irrigation accounts for 70% of the total water consumption, this is a logical place to start. The ultimate goal is to manage water at the level of an entire water catchment basin, accounting for surface and ground water and providing for the needs of all users, including the environment. WIN has developed a sensor/actuator network and a systems engineering approach to water management. The patented technology (commercialized as Total Channel Control™) is now being deployed in Australia’s largest irrigation district Goulburn Murray Water (GMW), consisting of 6800km of open irrigation canals servicing over 22,000 farms. The objective for the open canal system is to deliver water on demand (in as much this may be feasible) with maximal overall efficiency meeting the competing demands. We review the research work, including open questions, and discuss the WIN outcomes from a number of substantial pilot and commercial projects in Australia that have realized significant gains in either water efficiency or water productivity in irrigation.


23/04/2009 (14.00) [Lieu : Bât. Euler]
Robert BITMEAD (University of California, San Diego )
"Communications for Coordinated Control via Covariance Management"
Interaction between coordinating systems requires communication of information between the systems. Our aim is to quantify the required level and content of this communication. In the example considered, the interaction is developed as a non-collision constraint between vehicles and each vehicle solves a constrained Model Predictive Control (MPC) problem to determine its path - with the non-collision constraint being central. The extension of MPC from a full-state feedback strategy to one involving state estimates allows the adoption of the estimate covariance as a means to manage the communication requirements. This, in turn, admits a formulation of the communication resource assignment problem as a set of coupled Linear Matrix Inequalities, which link the control objectives, the external disturbance, and the communications.


21/04/2009 (14.00) [Lieu : Bât. Euler]
Michel JOURNEE (Ulg)
"Optimization methods for sparse principal component analysis"
Given an m-by-n matrix that stores m samples of n random variables, sparse principal component analysis (sparse PCA) is the problem of finding a linear combination of these variables that is sparse ( i.e., that involves as few as possible of the original variables) but that still captures as much as possible of the variance in the data. Sparsity is typically enforced for the sake of interpretability: components that are linear combinations of a small number of the original variables are easier to interpret. Sparse PCA has applications in virtually all areas of science, where large data sets need to be analyzed. This talk is organized in two parts. First, we formulate sparse PCA as the maximization of a convex function on a compact set and propose a simple gradient-method for this class of problems. We then consider convex relaxations of sparse PCA, which are optimization problems for positive semidefinite matrices of potentially large dimension. For the sake of numerical efficiency, we discuss an approach that rests on low-rank positive semidefinite matrices, which entail a quotient manifold structure.


31/03/2009 (14.00) [Lieu : Bât. Euler]
Marco DORIGO (IRIDIA/ULB)
"The Swarm-bot Experience in Swarm Robotics"
A swarm-bot is an artifact composed of a swarm of assembled robots called s-bots. The s-bots are mobile robots capable of connecting to, and disconnecting from, other s-bots. S-bots have relatively simple sensors and motors and limited compu- tational capabilities. In the swarm-bot form, the s-bots are attached to each other, forming a single robotic system that can move and change its shape. A swarm-bot can solve problems that cannot be solved by s-bots alone, and is par- ticularly robust to malfunctioning components. In the talk, after shortly describing the s-bots hardware and the methodology we followed to develop algorithms for their control, I will focus on the capabilities of the swarm-bot robotic system by showing video recordings of some of the many experiments we performed to study coordinated movement, path formation, self-assembly, collective transport, shape formation and reconfiguration, distributed fault detection and other collective behaviors.


24/03/2009 (14.00) [Lieu : Bât. Euler]
Emile SIMON (CESAME)
"LMI formulation for Optimal Control of coal fired Power Plants"
Thermal power plants are usually controlled by classical regulations (mostly PIDs), developed with a decentralised model, splitting the system in many different SISO loops. However, multiple providers and distributed generation are more likely to cause sudden and large changes of the load demand. Improving the performance is thus a commercial commodity of great importance in the current market’s state of privatised power industry. Therefore, we consider the solution of an optimal multivariable control. The objective pursued here was already studied in a previous work ["Performance Limitations arising in the Control of Power Plants" presented the 16/12/08 by Vincent Wertz], a key feature was the relation between the value of the tracking cost and the delay structure of the plant. Here a new method is chosen, the problem is cast into an LMI formulation. At first we develop the same trade-off results as previously found. But now the new formulation allows one to add constraints with other LMI's. Since in this paper the control energy is taken into account, a new LMI reducing the control input is developed. Without constraint, the coal flow input was too large. So a new LMI is introduced, allowing to reduce the norm of the optimal controller and thus the size of the input produced. This is shown to have a limited effect on the original objective, meaning a slight impact on the performance. We conclude from the results that the control is significantly improved, while having control inputs within their limits.


10/03/2009 (14.00) [Lieu : Bât. Euler]
Daniel COUTINHO (Pontifícia Universidade Católica do Rio Grande do Sul, Brazil.)
"An LMI approach for uncertain nonlinear systems". (slides)
In the last two decades, LMI-based techniques have been successfully applied to cope with uncertain linear systems in a wide diversity of applications such as robustness analysis, control and filter design, H2 and H-infinity design, control saturation and state-delayed systems. Together with the evolution of available LMI solvers, in terms of reliability, computational burden and easily of coding, many researchers have developed different LMI methodologies to deal with nonlinear systems. Recently, several different LMI-based techniques have been derived for nonlinear systems. One can divide these works in two main classes: (i) the linear-like representations, and (ii) the SOS (sum of squares) approach. The linear-like representations preserve the system state space representation and thus they can be applied for synthesis purposes. However, global stability conditions are of difficult handling. In contrast, the SOS-based approaches may handle global stability conditions but they are not suitable to control design. In this talk, the Differential-Algebraic Representation (DAR) framework is introduced to deal with a large class of nonlinear systems. Some recent results on stability analysis, performance and control synthesis are presented and a discussion is carried out on the advantages and shortcomings of the approach.


03/03/2009 (14.00) [Lieu : Bât. Euler]
Nicolas GILLIS (CORE)
"New variants of nonnegative matrix factorization for sparsity improvement and maximum biclique finding". (slides)
Nonnegative Matrix Factorization (NMF) is a data analysis technique which allows compression and interpretation of nonnegative data. It has been extensively and successfully applied in numerous applications, e.g. text mining, spectral data analysis, blind source separation, image processing, computational biology, graph clustering, recommendation systems, statistical models, etc. NMF consists in the factorization of a nonnegative matrix by the product of two low-rank nonnegative matrices. Because of the nonnegativity constraints, NMF is particularly well-suited to achieve decomposition into parts. In this talk, we first illustrate the usefulness of NMF with some application examples. We then present and analyze several algorithms to solve NMF. In this study, we introduce two new variants of NMF: Nonnegative Factorization (NF) and Nonnegative Matrix Underapproximation (NMU). NF is a generalization of NMF while NMU enables a recursive procedure for NMF and is particularly well-suited to achieve a better (sparser) part-based decomposition. Algorithms and applications for both NF and NMU are presented.Finally, NF and NMU are shown to be NP-hard using a reduction to a graph partitioning problem, namely the maximum-edge biclique problem (MBP). This reduction allows us design a new type of simple and efficient algorithm for MBP.


24/02/2009 (14.00) [Lieu : Bât. Euler]
Bruno DEHEZ (ELEC/LEI)
"Optimisation topologique : entre algorithmes meta-heuristiques et déterministes".
Les méthodes d'optimisation topologique permettent d'automatiser la conception de toutes sortes de dispositifs (structures mécaniques, dispositifs électromagnétiques, ...) grâce à l'action d'un algorithme d'optimisation dont l'objectif est de déterminer la meilleure manière de distribuer la matière dans un espace de design défini. Pratiquement, il s’agit de subdiviser cet espace de design en cellules chacune caractérisées par une série de paramètres fixant les propriétés (perméabilité magnétique, conductivité électrique et densité de courant dans le cas de dispositifs électromagnétiques) des matériaux les constituant. La recherche de la structure et de ses dimensions optimales est alors réalisée par un algorithme d’optimisation qui se charge de jouer avec les paramètres caractérisant les propriétés du matériau de chacune des cellules. Voilà maintenant un peu plus d'un an que nous avons initié dans notre laboratoire, en collaboration avec l'ENS de Cachan, des travaux de recherche sur ces méthodes d'optimisation topologique. Nous suivons deux approches distinctes. La première suppose que les matériaux de chacune des cellules peuvent varier de manière continue d'un matériau A vers un matériau B, ce qui permet de réaliser des études de sensibilités sur les paramètres, et de ce fait, d'utiliser des algorithmes d'optimisation déterministes. La seconde approche, par contre, ne considère pas que les matériaux peuvent varier de manière continue. Les algorithmes d'optimisation utilisés dans ce cas sont des algorithmes méta-heuristique pour lesquels le calcul des gradients n'est pas nécessaire. Bien que la convergence de ces algorithmes soit a priori plus lente que celle des algorithmes déterministes, ils ont de nombreux avantages, dont celui d'éviter plus facilement le piège de l'optimum local. Ce séminaire présentera l'état d'avancement de ces recherches ainsi que les premiers résultats obtenus avec les deux approches.


16/02/2009 (14.15) [Lieu : Barb.93]
Jean-Baptiste COULAUD (CESAME)
Thèse de doctorat "Singularités de Saut, Singularités de Platitude et Commande Optimale".
Ce document définit et décrit, dans le cadre des systèmes dynamiques à temps continu, un phénomène de discontinuité entre l'espace des trajectoires d'état et l'espace des trajectoires des sorties. Ce phénomène, que nous appellerons singularité de saut, apparaît dans certains systèmes dynamiques alors même que les fonctions des équations différentielles qui en définissent l'évolution sont parfaitement lisses. La thèse analyse le lien entre cette propriété et la propriété de platitude différentielle. Outre la compréhension structurelle des systèmes concernés que la notion de singularité de saut apporte, la thèse souligne aussi les implications sur la formulation de certains problèmes de commande optimale. Plusieurs exemples viennent illustrer cette problématique, notamment celui du robot mobile à plusieurs roues orientables qui a été le point de départ de l'analyse


12/02/2009 (16.30) [Lieu : Ch.-J. de La Vallée Poussin (cycl 01) ]
Stefaan VAES (K.U.L.)
Colloquium CESAME-MAPA "Algèbres de von Neumann et actions de groupes dénombrables préservant une mesure de probabilité".
Le sujet de cet exposé se trouve au carrefour de l'analyse fonctionnelle, la théorie ergodique et la théorie des groupes. En utilisant une construction due à Murray et von Neumann (1943), les groupes dénombrables et leurs actions ergodiques sur des espaces mesurés donnent lieu à des algèbres d'opérateurs sur un espace de Hilbert, que l'on appelle algèbres de von Neumann. Le but est d'expliquer la relation subtile entre une action de groupe et son algèbre de von Neumann : un théorème de Connes de 1976 dit que toutes les actions de tous les groupes moyennables mènent à une seule algèbre de von Neumann, tandis que les résultats récents de rigidité de Popa fournissent des actions de groupes telles que l'algèbre de von Neumann associée se souvient entièrement du groupe et de l'action.


10/02/2009 (14.00) [Lieu : Bât. Euler]
Olivier WHITE (School of Psychology, Bangor, UK)
"Responsibility Assignment in Redundant Systems".
Most movements are produced by the combined action of multiple actors or effectors. Such movements are highly redundant: each effector can substitute for the contribution of others. Recent research suggests that the brain distributes the work across different effectors such that signal-dependent noise and/or effort are minimized. When an error occurs during movement, however, another important problem arises: due to the redundancy, it is not clear from which effector the error originated. Which effector should carry the main responsibility for online error correction? For which effector should the motor command be updated in the next movement? Optimally, one would solve these questions based on separate principles: the effector that is better at correcting online errors should perform most of the correction, and the effector that is most likely responsible for the error, should adapt. We investigated how the brain distributes correction and adaptation across multiple effectors in a series of redundant tasks in which multiple effectors controlled the position of a single cursor. To probe online corrections and adaptation, the cursor was visually rotated by a random angle on each trial. We measured how much each effector corrected for the visual error by determining the difference between the initial and the overall movement direction. Participants showed a strong tendency to correct more with the left hand. We then estimated the learning rates for each of the involved effectors by fitting a state-space model to the initial movement directions. We found a clear correlation between the distribution of the online correction and the distribution of the adaptation. Thus, rather employing separate principles, our results clearly show the brain solves the responsibility assignment jointly for correction and adaptation. In other words, the effector that is found “guilty” of committing the error corrects more and adapts more.


16/12/2008 (14.00) [Lieu : Bât. Euler]
Vincent WERTZ (CESAME)
"Performance Limitations arising in the Control of Power Plants" (slides)
This talk presents results on performance limitations for direct fired coal power plants. A specific feature of this system is the existence of a very large input delay between one of the inputs, namely coal flow, and the two outputs, load and vapour pressure. This problem motivates the main theoretical question addressed in this paper: To examine tracking performance limitations in one process variable when another process variable is constrained. Our main result makes explicit the performance trade-off between the two conflicting objectives, and also links the achievable performance to the delay structure of the plant. These results give insights into the benefits of MIMO control for power plants and into the necessary trade-off between fast tracking of load step changes and the need for minimizing the variations of the vapour pressure around its nominal value. The results provide a benchmark against which practical controller designs for power plants can be assessed.


9/12/2008 (14.00) [Lieu : Bât. Euler]
Frédéric CREVECOEUR (CESAME)
"Trajectory planning of vertical pointing movements: using gravity to move the arm" (slides)
This study aims at better understanding how the central nervous system (CNS) integrates the action of the gravitational torque on the upper limb when we perform a vertical movement such as replacing a cup on a shelf. Particular kinematic properties of vertical pointing movements have provided evidence for an integration of the action of gravity in the motor plan. However, it remains unclear how the CNS achieves this integration and optimal control approaches considering minimum joint torque input failed to explain the kinematic properties observed in normal gravity condition. In this experiment, we asked subjects to perform vertical pointing movements in different gravitational conditions, normal (1g), hyper (1.8g) and micro gravity (~0g). We used a model that considers minimal motoneuron discharge to test the hypothesis that gravity is integrated in an energy-optimal way. The observed behaviors are consistent with the model's predictions although the strategies vary across subjects, movement's direction and gravity condition. The analysis of subjects' strategies permit to better understand the nature of mechanism underlying adaptation of motor control to altered gravity condition.


3/12/2008 (16.00) [Lieu : Aud. BARB92, Place Sainte Barbe]
Erik DEMAINE (MIT, U.S.A.)
CHAIRE FRANQUI - (Teoretical) Computer Science is Everywhere
Theoretical computer science, and the algorithmic way of thinking, transcends our traditional boundaries. I believe that algorithms are relevant to every discipline of study, and will give eclectic examples from the arts and sciences to business and society. The examples span the spectrum from serious topics like protein folding and decoding Inka khipu to fun topics like juggling and magic.


27/11/2008 (16.30) [Lieu : Auditoire Ch.-J. de La Vallée Poussin (cycl 01) ]
Frédéric BOURGEOIS (ULB)
Colloquium CESAME-MAPA "Un survol de l'homologie de contact"
L'homologie de contact est un invariant pour les structures de contact, dont la définition est basée sur les courbes holomorphes. Dans cet exposé, on esquissera la construction de l'homologie de contact, puis on illustrera certaines de ses applications et on discutera de certaines de ses propriétés.


25/11/2008 (14.00) [Lieu : Bât. Euler]
Michel PERRIER (Ecole Polytechnique de Montréal)
"Microbial Fuel Cells : Principles and Optimization by Multi Unit Approach" (slides)
Multi-unit optimization is an extremum seeking method (real-time steady state optimization of a dynamic system by controlling the gradient), where the gradient is obtained using finite difference of the outputs of multiple identical units driven with different inputs. The major limitation of this approach is that the convergence is established only when units are identical. This talk addresses the convergence of the multi-unit scheme with non-identical units. With a simple quadratic approximation of the objective function, it is shown that the scheme converges only when the difference between the inputs is either larger than or in a direction opposite to that of the difference between their optimum values. Also, it is shown that the best value for the difference between the inputs is not zero, but determined by the deviation between the units. The multi-unit algorithm is applied to the maximization of power production in a stack of two continuous flow microbial fuel cells (MFCs) to maximize power output by manipulating external resistances of two air­cathode membraneless MFCs. The experiment demonstrated fast convergence toward optimal external resistance and algorithm stability during external perturbations (e.g. temperature variations). Rate of the algorithm convergence was much faster than in traditional maximum power point tracking algorithms (MPPT), which are based on temporal perturbations.


18/11/2008 (14.00) [Lieu : Bât. Euler]
Alain SARLETTE (ULg)
"Coordination on nonlinear spaces" (slides)
Recently, numerous authors have studied algorithms for interacting individual agents that lead to a coordinated behavior of the overall swarm. Mostly, the agents are assumed to evolve on a vector space.
This talk extends the theoretical study of coordination to nonlinear but highly symmetric spaces, like the spheres or Lie groups. The focus is on consequences of the geometry and associated symmetries of the setting. It shows that new fundamental problems appear in this context, proposes appropriate definitions and various control methods to circumvent undesired phenomena. Two main "coordination" tasks are considered: (i) achieving a particular configuration of the agents and (ii) performing "coordinated motions".


4/11/2008 (14.00) [Lieu : Euler]
Cosmin LAZAR (University Reims, France)
"Unsupervised methods for multivariate data analysis" (slides)
Multivariate data analysis gathers efficient tools for extracting reliable information in order to comprehend the phenomenon in study. Gathering data into groups or classes according to some similarity criteria is an essential step in the analysis. Intrinsic dimension or dimension reduction of multivariate data, the choice of the similarity criterion, cluster validation are problems which still let open questions.

This work tries to make a step further concerning two of the problems mentioned above: the choice of the similarity measure for data clustering and the dimension reduction of multivariate data. The choice of the similarity measure for data clustering is investigated from the concentration phenomenon of metrics point of view. Non Euclidean metrics are tested as alternative to the classical Euclidian distance as similarity measure. We tested if less concentrated metrics are more discriminative for multivariate data clustering. We also proposed indices which take into account the inter-classes distance (e.g. Davies-Bouldin index) in order to find the optimal metric when the classes are supposed to be Gaussian.

Blind Source Separation (BSS) methods are also investigated for dimension reduction of multivariate data. A BSS method based on a geometrical interpretation of the linear mixing model is proposed. BSS methods which take into account application constraints are used for dimension reduction in two different applications of multivariate imaging. These methods allow the extraction of meaningful factors from the whole data set; they also allow reducing the complexity and the computing time of the clustering algorithms which are used further in analysis.


28/10/2008 (14.00) [Lieu : Euler]
Tzvetan IVANOV (CESAME)
"Applications of Real Rational Modules in System Identification" (slides)
We introduces a real rational module framework in the context of Prediction Error Identification using Box-Jenkins model structures. This module framework, which can easily be extended to other model structures, allows us to solve and/or extend a number of problems related to the computation of error norms that arise in system identification. Our main contribution to system identification is an extension of the asymptotic variance formulas for Box-Jenkins models derived by Ninness and Hjalmarsson to asymptotic auto-covariance with respect to frequency. This auto-covariance is in fact the integral kernel reproducing the tangent space of the model manifold at the true (or fixed) system w.r.t. the Fisher-information metric. We show how to exploit the fact that the tangent space forms a rational module in order to quantify this kernel in terms of the poles and zeros of the true system and the input spectrum.


21/10/2008 (14.00) [Lieu : Euler ]
Jean-Michel CORON (Université Pierre et Marie Curie, Laboratoire Jacques-Louis Lions, France)
"Quelques méthodes pour étudier la contrôlabilité des équations aux dérivées partielles non linéaires." (slides)
Quand on veut regarder la contrôlabilité locale autour d'un point d'équilibre d'un système de contrôle non linéaire modélisé par une équation aux dérivées partielles, on commence, comme en dimension finie, par regarder la contrôlabilité du linéarisé. Mais, déjà, à cause d'un problème de ``pertes de dérivées'', des difficulté apparaissent souvent pour déduire de la contrôlabilité du linéarisé la contrôlabilité locale du non linéaire. De plus si le linéarisé n'est pas contrôlable on ne peut rien déduire pour la contrôlabilité du système non linéaire. Les méthodes utilisant les crochets de Lie itérés pour traiter ce dernier cas en dimension finie marchent mal pour les équations aux dérivées partielles. On présentera d'autres méthodes utiles pour traiter ce cas (méthode du retour, déformation quasi-statiques, développement en série). On donnera des exemples récemment traités par ces méthodes.


14/10/2008 (14.00) [Lieu : Euler ]
Didier HENRION (LAAS-CNRS, University of Toulouse, France, and Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic )
"Plane geometry and convexity of polynomial stability regions" (slides)
The set of controllers stabilizing a linear system is generally non-convex in the parameter space. In the case of two-parameter controller design (e.g. PI control or static output feedback with one input and two outputs), we observe however that quite often for benchmark problem instances, the set of stabilizing controllers seems to be convex. In this note we use elementary techniques from real algebraic geometry (resultants and B´ezoutian matrices) to explain this phenomenon. As a byproduct, we derive a convex linear matrix inequality (LMI) formulation of two-parameter fixed-order controller design problem, when possible.


8/10/2008 (11.45) [Lieu : Place Croix du Sud (SUD19)]
John N. TSITSIKLIS (MIT, USA)
D.H.C. - "Decentralized Detection"
Doctor Honoris Causa - Wednesday, October 8th


30/09/2008 (14.00) [Lieu : Bât. Euler]
Robert DAVID (Cesame)
"Modélisation dynamique et commande des procédés d'épuration biologique des eaux usées à boues activées" (slides)
Le procédé d'épuration par boues activées est un traitement biochimique qui dégrade les éléments carbonés et azotés présents dans les eaux usées. Cet exposé met d'abord l'accent sur la modélisation de la décantation des boues, une étape primordiale du procédé, avant de considérer un modèle du procédé entier. La réduction et la linéarisation de celui-ci permet alors une synthèse de commande robuste par retour d'état qui régule le taux d'oxygène et les concentrations en sortie des polluants.


23/09/2008 (14.00) [Lieu : Bât. Euler]
Huseyin AKCAY (Anadolu University, Turkey)
"A Subspace-Based Method for Solving Lagrange--Sylvester Interpolation" (slides)
In this talk, we study the Lagrange--Sylvester interpolation of rational matrix functions which are analytic at infinity, and propose a new interpolation algorithm based on the recent subspace-based identification methods. The proposed algorithm is numerically efficient and delivers a minimal interpolant in state-space form. The solvability condition for the subspace-based algorithm is particularly simple and depends only on the total multiplicity of the interpolation nodes. As an application, we consider subspace-based system identification with interpolation constraints, which arises, for example, in the identification of continuous-time systems with a given relative degree.


16/09/2008 (14.00) [Lieu : Bât. Euler]
Brian ANDERSON (Australian National University)
"Computational Approaches to H∞ problems and Differential Games" (slides)
Linear H∞ problems require the solution of Riccati equations with an indefinite quadratic term, which can be computationally challenging. For H2 Riccati equations, numerical problems can often be overcome using the Kleinman iteration, involving a sequence of Lyapunov equations. While this is not directly possible for Riccati equations with an indefinite quadratic term, an iteration involving a sequence of H2 equations is possible which allows a double iteration involving Lyapunov equations.

These ideas are given a game theory interpretation, which then suggests an extension of the results that will also be exposed; the extension indicates how a Hamilton-Jacobi-Bellman-Isaacs equation for a class of nonlinear game theory problems can be solved by iterating of Hamilton-Jacobi-Bellman equations, each of which in turn can be solved by iteration using linear partial differential equations.


30/05/2008 (11.00) [Lieu : Bât. Euler ]
Abdellah BENZAOUIA (EACPI, University Cadi Ayyad, )
Stability and Control Synthesis of Switching Systems Subject to Actuator Saturation
This paper presents sufficient conditions for the stabilization of switching discrete-time linear systems subject to actuator saturations. These conditions are obtained by using successively state and output feedback control laws. The obtained results are formulated in terms of LMI's. Three sets of LMI's are presented for output feedback case. A numerical example is used to illustrate the technique by using a linear optimization problem subject to LMI constraints. Key-words: Switching systems, Actuator saturations, invariant sets, Lyapunov functions, LMI's.


20/05/2008 (14.00) [Lieu : Bât. Euler ]
Pierre Daye ( CESAME )
Adding head to eye movements: what are the implications for visual tracking?
In everyday life, humans make combined eye-head movements to reorient their gaze (Gaze = Eye movement with respect to the head + head movement with respect to the torso). Combined eye-head movements are necessary when the targets of interest are outside the so-called "oculomotor range". In this seminar, we will first present the main characteristics of combined eye-head movements. We will then show and discuss results from a 2D gaze orientation task to memorized targets.


24/04/2008 (16.30) [Lieu : Bât. Euler ]
Pascal Lambrechts ( UCL )
La conjecture de Poincaré et la classification des 3-variétés.
D'après le magazine Science, l'évènement scientifique de l'année 2006 fut la résolution de la conjecture de Poincaré par Grigori Perelman. Dans cet exposé j'expliquerai cette conjecture et des idées élémentaires de sa preuve. Mais j'expliquerai aussi que Perelman a en fait démontré beaucoup plus que la conjecture de Poincaré puisqu'il semble avoir achevé un formidable programme de recherche initié par Thurston à la fin des années 1970. Ce programme proposait une classification des variétés de dimension 3 grâce à l'existence de géométries homogènes sur ces variétés, en analogie avec un phénomène analogue en dimension 2 bien connu selon lequel toute surface compacte orientable admet une géométrie à courbure constante.


22/04/2008 (14.00) [Lieu : Bât. Euler ]
Jérémy Badler ( CESAME )
Interactions Between Explicit and Implicit Timing
When stimuli are temporally regular, their prediction depends on an implicit representation of elapsed time. However, it is unclear whether the underlying timing mechanism is separable from a conscious, explicit perception of time. To address the question, human subjects had to predict the duration of an interval either explicitly, with a button press, or implicitly, by accurately tracking with the eyes a suddenly-moving target. Performing the tasks concurrently yielded a delay in response times for both modalities compared to single-task trials. This suggests that implicit and explicit time representations are partially independent, and invoking them simultaneously increases the attentional load on the brain.


15/04/2008 (14.00) [Lieu : Bât. Euler ]
Cristobald de KERCHOVE ( CESAME )
Iterative filtering for a dynamical reputation system
I introduces a novel iterative method that assigns a reputation to $n+m$ items: n raters and m objects. Each rater evaluates a subset of objects leading to a nxm evaluation matrix with a certain sparsity pattern. From this evaluation matrix we give a nonlinear formula to define the reputation of raters and objects. I also provide an iterative algorithm that superlinearly converges to the unique vector of reputations and this for any evaluation matrix. In contrast to classical outliers detection, no evaluation is discarded in this method but each one is taken into account with different weights for the reputation of the objects. The complexity of one iteration step is linear in the number of evaluations, making our algorithm efficient for large data set. Experiments show good robustness of the reputation of the objects against cheaters and spammers and good detection properties of cheaters and spammers.


08/04/2008 (14.00) [Lieu : Bât. Euler ]
Gianna Del Corso ( University of Pisa )
Evaluating scientific products by means of citation-based models: a first analysis and validation
Some integrated models for ranking scientific publications together with authors and journals are presented and analyzed. The models rely on certain adiacency matrices obtained from the relations of citation, authorship and publication, which concurr to forming a suitable irreducible stochastic matrix whose Perron vector provides the ranking. Some perturbation theorems concerning the Perron vector of nonnegative irreducible matrices are proved. These theoretical results provide a validation of the consistency and effectiveness of our models. Several paradigmatic examples are reported together with some results obtained on a real set of data.


18/03/2008 [Lieu : Heeze, The Netherlands ]
27th Benelux Meeting on Systems and Control - NO SEMINAR !!


11/03/2008 (14.00) [Lieu : Bât. Euler ]
Marcelo MOISAN ( INRIA-Sophia Antipolis )
Near optimal and reverse time interval observers. Application to uncertain bioreactors
In this seminar we propose a new interval observer for the estimation of unmeasured variables of uncertain bioreactors. The observer is based on a class of bounded error observers, that make use of a loose approximation of the bacterial kinetics. We first show how to generate guaranteed upper and lower bounds on the unknown state, provided that known intervals for the initial condition and the uncertainties are available. An optimality criterion is then introduced, leading to the definition of an optimal observer. We show that this criterion allows to determine a set of gains which generates the best estimates. We extend the observer design introducing reverse time interval observers. We show that running observers in reverse time improves dramatically the convergence rate, through a discontinuous interval estimation procedure. The method is applied to the estimation of the total biomass of an industrial wastewater treatment plant, demonstrating its efficiency.


04/03/2008 (14.00) [Lieu : Bât. Euler ]
Chafik SAMIR ( Université de Sciences et Technologies de Lille USTL/CESAME )
A framework of calculus on facial surfaces : Analysis representation and comparison
There is an increasing interest in analyzing shapes of facial surfaces with many applications including biometrics, facial surgery, video communications, and 3D animation. This interest is fuelled by the advent of different types of scanners that can provide high-resolution measurements of both geometry and texture of human facial surfaces. One general goal in this work is to develop computational tools for analyzing 3D face data. In particular, we are interested in comparing the shapes of facial surfaces. Such a tool can be used to recognize human beings according to their facial shapes, to measure changes in a facial shape due to a surgery, or to study/capture the variations in facial shapes during conversations and expressions of emotions. Additionally, another subproblem addressed in here is to find an optimal deformation of one surface into another, under some chosen criterion. These deformations can be useful in defining summary statistics of a collection of faces. For example, we define and compute an it average face (Karcher mean) of a set of people, or for one person under different facial expressions. Efficient tools for understanding and studying variability in facial shapes are of great importance in our biometric-oriented society, our method using planar shapes is tested on FRGC v1.0 database, and obtained results are very promising.


26/02/2008 (14.00) [Lieu : Bât. Euler ]
Nadia BENIICH (University of El Jadida/CESAME )
Régulation de la température du réacteur tubulaire à l'intérieur du réacteur chimique tubulaire.
On s'intéresse dans cet exposé à la régulation de la température des réacteurs tubulaires exothermiques à dispersion axiale et piston dont la dynamique est décrite par des équations aux dérivées partielles non linéaires, en utilisant des commandes aux bords soumises à des contraintes. Nous proposons quelques contrôleurs adaptatifs qui permettent de conduire la température du réacteur à l'intérieur d’un voisinage d'un profil de référence de rayon ? arbitraire et suffisamment petit.


21/02/2008 (16.00) [Lieu : BARB93 ]
Laure NINOVE ( CESAME )
JEUDI !! THESE DE DOCTORAT "Dominant vectors of nonnegative matrices. Application to information extraction in large graphs".
Objects such as documents, people, words or utilities, that are related in some way, for instance by citations, friendship, appearance in definitions or physical connections, may be conveniently represented using graphs or networks. An increasing number of such relational databases, as for instance the World Wide Web, digital libraries, social networking web sites or phone calls logs, are available. Relevant information may be hidden in these networks. A user may for instance need to get authority web pages on a particular topic or a list of similar documents from a digital library, or to determine communities of friends from a social networking site or a phone calls log. Unfortunately, extracting this information may not be easy. This thesis is devoted to the study of problems related to information extraction in large graphs with the help of dominant vectors of nonnegative matrices. The graph structure is indeed very useful to retrieve information from a relational database. The correspondence between nonnegative matrices and graphs makes Perron-Frobenius methods a powerful tool for the analysis of networks. In a first part, we analyze the fixed points of a normalized affine iteration used by a database matching algorithm. Then, we consider questions related to PageRank, a ranking method of the web pages based on a random surfer model and used by the well known web search engine Google. In a second part, we study optimal linkage strategies for a web master who wants to maximize the average PageRank score of a web site. Finally, the third part is devoted to the study of a nonlinear variant of PageRank. The simple model that we propose takes into account the mutual influence between web ranking and web surfing.


14/02/2008 (14.00) [Lieu : BARB93 ]
Julien HENDRICKX ( CESAME )
JEUDI!! THESE DE DOCTORAT "Graphs and networks for the analysis of autonomous agent systems".


05/02/2008 (14.00) [Lieu : Bât. Euler ]
Paul BELL (University of Liverpool/CESAME )
Computability in Linear Dynamical Systems
We shall present some current results of computational problems on matrix semigroups and low dimensional dynamical systems with an emphasis on reachability and recurrence type problems. In such problems we are given a finite set of matrices over a semiring and wish to decide whether certain properties of the semigroup generated by this set are algorithmically decidable or not. These problems very often have interpretations in terms of linear dynamical systems and we shall highlight the connection between these two subjects. We shall also see some matrix interpretations of Skolem's problem concerning linear recurrent sequences. Open problems in these areas and possibilities for future work will also be presented.


29/01/2008 (14.00) [Lieu : Bât. Euler ]
Maher MOAKHER ( Ecole Nationale d'Ingenieurs de Tunis )
A Riemannian Framework for the Regularization of Symmetric Positive-Definite Tensor Fields
In this talk we present a Riemannian framework for the smoothing of data that are constrained to live in the space of symmetric positive-definite matrices. We start by giving the differential geometry of this space. We then use the harmonic map and minimal immersion theories to construct three flows that drive a noisy field of symmetric positive-definite data into a smooth one. The harmonic map flow is the equivalent of the heat flow or isotropic linear diffusion which smooths data everywhere. A modification of the harmonic flow leads to a Perona-Malik like flow which is a selective smoother that preserves edges. The minimal immersion flow gives rise to a nonlinear system of coupled diffusion equations with anisotropic diffusivity. Some preliminary numerical results are presented for synthetic diffusion tensor MRI data.


08/01/2008 (14.00) [Lieu : Bât. Euler ]
Ignasi COS-AGUILERA (HWNI, UC Berkeley, USA)
An Introduction to Dynamic Encoding
The Brain Machine Interface paradigm relies on cortical adaptivity to generate and control movement of an external device. In particular, previous experiments have demonstrated the validity of motor brain areas for controlling a cursor on a computer screen with the help of visual feedback. However, these approaches lack control of every parameter involved in the motion of a physical object, the parameters controlling the dynamics of the movement: inertia, stiffness and viscosity. In a complementary fashion to this main stream, we argue that decoding of those parameters from the cortex is the most natural procedure to control a prosthetic device. Eventually, dynamic control should be the one requiring the least effort to adapt to. From this perspective, we have been studying the procedures of control of movement within force-field conditions via psychophysical experiments, aiming at understanding the adaptation process of the brain to situations requiring accountancy of the dynamics of motion. The experiments demonstrate different strategies to cope with different force-field conditions, suggesting that once adaptation is complete, movement is controlled via modulation of impedance. Stiffness, viscosity and intertia are optimized according to the position and type of reach. We present some of these experiments together with a prospect of future research.



| 23/06/2006 |